Kannada word sense disambiguation using semantic relations

Abstract

In this paper, a method is proposed to use the semantic relations of words to solve Word Sense Disambiguation (WSD) for the Kannada language. Semantic relationships like hypernymy and hyponymy are used in finding the overlaps between two concepts. Integrating the idea proposed in this paper with original Lesk's technique improves the accuracy of the Lesk's algorithm. Indo-WordNet is used as the lexical resource to gain the knowledge about the semantic relations of words and this knowledge is exploited to disambiguate the target word of the Kannada language. Since the approach used in this paper is unsupervised, so it can be scalable to other languages also.